pith. sign in

arxiv: 2407.13059 · v2 · pith:C6YMIQK5new · submitted 2024-05-25 · 💻 cs.CY · cs.AI· cs.ET

Prioritizing High-Consequence Biological Capabilities in Evaluations of Artificial Intelligence Models

Jaspreet Pannu , Doni Bloomfield , Alex Zhu , Robert MacKnight , Gabe Gomes , Anita Cicero , Thomas V. Inglesby This is my paper

Pith reviewed 2026-05-24 01:08 UTC · model grok-4.3

classification 💻 cs.CY cs.AIcs.ET
keywords AI safetybiosecuritydual-use researchbiological capabilitiesmodel evaluationpandemic risksAI governancepre-deployment assessment
0
0 comments X

The pith

AI model evaluations should prioritize high-consequence biological risks like pandemics and assess them before deployment.

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The paper argues that decades of work identifying and mitigating dual-use research in the life sciences—experiments that can both advance knowledge and risk causing large-scale harm—can guide evaluations of AI models that possess biological capabilities. It claims these evaluations must focus first on risks capable of causing pandemics or similar public harm, with assessments completed prior to any model release so that biosafety or biosecurity steps can be taken if needed. A sympathetic reader would care because AI advances could amplify existing pathways to engineered pathogens, and the life-sciences precedent offers a tested method for distinguishing dangerous capabilities from beneficial ones. The authors state that identifying the AI capabilities of greatest concern is required to build targeted safety methods, protect against accident and misuse, and keep beneficial applications open. This transfers established dual-use practices to the newer domain of AI without requiring entirely new frameworks.

Core claim

Experience and study by scientists and policy professionals of dual-use capabilities in the life sciences can inform risk evaluations of AI models with biological capabilities. AI model evaluations should prioritize addressing high-consequence risks that could cause large-scale harm to the public, such as pandemics, and these risks should be evaluated prior to model deployment so as to allow potential biosafety and/or biosecurity measures. Identifying which AI capabilities pose the greatest biosecurity and biosafety concerns is necessary in order to establish targeted AI safety evaluation methods, secure these tools against accident and misuse, and avoid impeding immense potential benefits.

What carries the argument

Dual-use identification and mitigation approaches from the life sciences, transferred to evaluate AI models for biological capabilities.

If this is right

  • Targeted AI safety evaluation methods can be established for the capabilities of greatest concern.
  • AI tools can be secured against accident and misuse through pre-deployment biosafety measures.
  • Immense potential benefits of AI in biology can proceed without broad impediments.
  • Biosecurity measures can be applied where high-consequence risks are identified.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • The same prioritization logic could be tested on AI capabilities in other high-risk domains such as chemistry.
  • Developers might incorporate these pre-deployment checks into standard release pipelines for frontier models.
  • International standards bodies could adopt the dual-use criteria as a baseline for AI governance.
  • Early integration of these evaluations during model training could reduce the cost of later mitigation.

Load-bearing premise

Dual-use identification and mitigation approaches developed in the life sciences can be directly and effectively transferred to AI models to identify and mitigate high-consequence biological capabilities without excessive false positives or stifling beneficial applications.

What would settle it

An empirical test showing that life-sciences dual-use criteria applied to AI models produce widespread false positives that block safe models or miss genuine high-consequence biological capabilities.

Figures

Figures reproduced from arXiv: 2407.13059 by Alex Zhu, Anita Cicero, Doni Bloomfield, Gabe Gomes, Jaspreet Pannu, Robert MacKnight, Thomas V. Inglesby.

Figure 1
Figure 1. Figure 1: Biosecurity evaluation development should follow a distinct process. This figure is a simple schematic demonstrating that distinct evaluative methods will be needed for different types of AI models. High-consequence capabilities of concern should be translated into model type-specific evaluation methods that are linked to pre-specified thresholds and risk mitigation actions. Some companies have stated that… view at source ↗
read the original abstract

As a result of rapidly accelerating AI capabilities, over the past year, national governments and multinational bodies have announced efforts to address safety, security and ethics issues related to AI models. One high priority among these efforts is the mitigation of misuse of AI models. Many biologists have for decades sought to reduce the risks of scientific research that could lead, through accident or misuse, to high-consequence disease outbreaks. Scientists have carefully considered what types of life sciences research have the potential for both benefit and risk (dual-use), especially as scientific advances have accelerated our ability to engineer organisms and create novel variants of pathogens. Here we describe how previous experience and study by scientists and policy professionals of dual-use capabilities in the life sciences can inform risk evaluations of AI models with biological capabilities. We argue that AI model evaluations should prioritize addressing high-consequence risks (those that could cause large-scale harm to the public, such as pandemics), and that these risks should be evaluated prior to model deployment so as to allow potential biosafety and/or biosecurity measures. Scientists' experience with identifying and mitigating dual-use biological risks can help inform new approaches to evaluating biological AI models. Identifying which AI capabilities post the greatest biosecurity and biosafety concerns is necessary in order to establish targeted AI safety evaluation methods, secure these tools against accident and misuse, and avoid impeding immense potential benefits.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Referee Report

0 major / 2 minor

Summary. The paper claims that evaluations of AI models should prioritize high-consequence biological risks (those with potential for large-scale public harm such as pandemics) by drawing on established dual-use research practices from the life sciences; it recommends conducting such evaluations prior to model deployment to enable biosafety and biosecurity measures, while arguing that life-sciences experience can inform targeted AI evaluation methods without impeding beneficial applications.

Significance. If the central recommendation holds, the manuscript provides a policy framework that focuses AI biosecurity efforts on the most serious risks by explicit analogy to decades of life-sciences dual-use precedents; this is a strength because the paper draws on external, long-established literature rather than self-referential results, offering a directional recommendation for prioritization that could help avoid both excessive false positives and under-regulation of high-stakes capabilities.

minor comments (2)
  1. [Abstract] The abstract and introduction could more precisely delineate the class of AI models under discussion (e.g., general-purpose LLMs versus specialized biological design tools) to sharpen the scope of the recommended evaluations.
  2. A short table or bullet list summarizing the key dual-use criteria from the life-sciences literature (with citations) would improve readability when the analogy is applied to AI capabilities.

Simulated Author's Rebuttal

0 responses · 0 unresolved

We thank the referee for their positive assessment and recommendation to accept the manuscript. The referee's summary correctly identifies the paper's core argument that AI biological capability evaluations should draw on established life-sciences dual-use frameworks to prioritize high-consequence risks and enable pre-deployment biosafety measures.

Circularity Check

0 steps flagged

No significant circularity

full rationale

The paper advances a policy recommendation to prioritize high-consequence biological risks in AI evaluations by drawing on decades of external life-sciences dual-use research and established biosafety practices. No equations, fitted parameters, predictions, or derivations are presented whose validity reduces to self-citation chains, self-definitional constructs, or inputs renamed as outputs. The central argument is an independent directional recommendation grounded in external precedent rather than any load-bearing internal reduction.

Axiom & Free-Parameter Ledger

0 free parameters · 1 axioms · 0 invented entities

The central claim rests on the domain assumption that life sciences dual-use frameworks transfer effectively to AI; no free parameters or new entities are introduced.

axioms (1)
  • domain assumption Dual-use research experience in the life sciences provides a valid and transferable framework for identifying and prioritizing high-consequence risks in AI biological capabilities.
    This premise is invoked as the basis for the entire recommendation in the abstract and paper.

pith-pipeline@v0.9.0 · 5795 in / 1240 out tokens · 33101 ms · 2026-05-24T01:08:12.506394+00:00 · methodology

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Reference graph

Works this paper leans on

62 extracted references · 62 canonical work pages · 4 internal anchors

  1. [1]

    Nat Mach Intell

    AI protein shake-up. Nat Mach Intell. 2024 Feb;6(2):121–121

    work page 2024

  2. [2]

    Maug N. Epoch. 2024 [cited 2024 Mar 27]. Biological Sequence Models in the Context of the AI Directives. Available from: https://epochai.org/blog/biological-sequence-models-in-the-context-of-the- ai-directives

    work page 2024

  3. [3]

    Protein design meets biosecurity

    Baker D, Church G. Protein design meets biosecurity. Science. 2024 Jan 26;383(6681):349–349

    work page 2024

  4. [4]

    Autonomous chemical research with large language models

    Boiko DA, MacKnight R, Kline B, Gomes G. Autonomous chemical research with large language models. Nature. 2023 Dec;624(7992):570–8

    work page 2023

  5. [5]

    arXiv.org

    Kapoor S, Bommasani R, Klyman K, Longpre S, Ramaswami A, Cihon P, et al. arXiv.org. 2024 [cited 2024 Mar 27]. On the Societal Impact of Open Foundation Models. Available from: https://arxiv.org/abs/2403.07918v1

    work page arXiv 2024

  6. [6]

    [cited 2024 Mar 27]

    Building an early warning system for LLM-aided biological threat creation [Internet]. [cited 2024 Mar 27]. Available from: https://openai.com/research/building-an-early-warning-system-for-llm-aided- biological-threat-creation

    work page 2024

  7. [7]

    The Operational Risks of AI in Large-Scale Biological Attacks: A Red-Team Approach [Internet]

    Mouton CA, Lucas C, Guest E. The Operational Risks of AI in Large-Scale Biological Attacks: A Red-Team Approach [Internet]. RAND Corporation; 2023 Oct [cited 2024 Mar 27]. Available from: https://www.rand.org/pubs/research_reports/RRA2977-1.html

    work page 2023

  8. [8]

    Cornell University Press

    Barriers to Bioweapons by Sonia Ben Ouagrham-Gormley | Hardcover [Internet]. Cornell University Press. [cited 2024 Mar 27]. Available from: https://www.cornellpress.cornell.edu/book/9780801452888/barriers-to-bioweapons/

    work page arXiv 2024

  9. [9]

    Phantom Menace or Looming Danger? [Internet]

    Vogel KM. Phantom Menace or Looming Danger? [Internet]. Johns Hopkins University Press; 2012 [cited 2024 Mar 27]. Available from: https://www.press.jhu.edu/books/title/10403/phantom-menace- or-looming-danger

    work page 2012

  10. [10]

    Opinion | What if Dario Amodei Is Right About A.I.? The New York Times [Internet]

    Show’ ‘The Ezra Klein. Opinion | What if Dario Amodei Is Right About A.I.? The New York Times [Internet]. 2024 Apr 12 [cited 2024 May 6]; Available from: https://www.nytimes.com/2024/04/12/opinion/ezra-klein-podcast-dario-amodei.html

    work page 2024

  11. [11]

    Diving deep into OpenAI’s new study on LLM’s and bioweapons [Internet]

    Marcus G. Diving deep into OpenAI’s new study on LLM’s and bioweapons [Internet]. Marcus on AI. 2024 [cited 2024 Mar 27]. Available from: https://garymarcus.substack.com/p/when-looked-at- carefully-openais

    work page 2024

  12. [12]

    AI’s bioterrorism potential should not be ruled out [Internet]

    Ahuja A. AI’s bioterrorism potential should not be ruled out [Internet]. 2024 [cited 2024 Mar 27]. Available from: https://www.ft.com/content/e2a28b73-9831-4e7e-be7c-a599d2498f24

    work page 2024

  13. [13]

    How to better research the possible threats posed by AI-driven misuse of biology [Internet]

    Goudarzi S. How to better research the possible threats posed by AI-driven misuse of biology [Internet]. Bulletin of the Atomic Scientists. 2024 [cited 2024 Mar 27]. Available from: https://thebulletin.org/2024/03/how-to-better-research-the-possible-threats-posed-by-ai-driven-misuse- of-biology/

    work page 2024

  14. [14]

    Artificial intelligence and biological misuse: Differentiating risks of language models and biological design tools [Internet]

    Sandbrink JB. Artificial intelligence and biological misuse: Differentiating risks of language models and biological design tools [Internet]. arXiv; 2023 [cited 2024 Apr 14]. Available from: http://arxiv.org/abs/2306.13952

    work page arXiv 2023

  15. [15]

    [cited 2024 Jun 21]

    End 2 End AI Molecule Design – We generate diverse de novo protein sequences from just a text description of the desired properties by Mol.E , a state-of-the-art ML model [Internet]. [cited 2024 Jun 21]. Available from: https://310.ai/

    work page 2024

  16. [16]

    [cited 2024 Jun 21]

    FutureHouse [Internet]. [cited 2024 Jun 21]. Available from: https://www.futurehouse.org/

    work page 2024

  17. [17]

    [cited 2024 Jun 21]

    SAM.gov [Internet]. [cited 2024 Jun 21]. Available from: https://sam.gov/opp/dd906dc45ee347d5a0c29d980cf67dcc/view

    work page 2024

  18. [18]

    [cited 2024 Mar 27]

    Responsible AI x Biodesign [Internet]. [cited 2024 Mar 27]. Responsible AI x Biodesign. Available from: https://responsiblebiodesign.ai/

    work page 2024

  19. [19]

    Self-driving laboratories to autonomously navigate the protein fitness landscape

    Rapp JT, Bremer BJ, Romero PA. Self-driving laboratories to autonomously navigate the protein fitness landscape. Nat Chem Eng. 2024 Jan;1(1):97–107

    work page 2024

  20. [20]

    In vitro continuous protein evolution empowered by machine learning and automation

    Yu T, Boob AG, Singh N, Su Y, Zhao H. In vitro continuous protein evolution empowered by machine learning and automation. Cell Syst. 2023 Aug 16;14(8):633–44

    work page 2023

  21. [21]

    Laboratories in the cloud [Internet]

    Field M. Laboratories in the cloud [Internet]. Bulletin of the Atomic Scientists. 2019 [cited 2024 Mar 27]. Available from: https://thebulletin.org/2019/07/laboratories-in-the-cloud/

    work page 2019

  22. [22]

    Federation of American Scientists

    Bio x AI: Policy Recommendations for a New Frontier [Internet]. Federation of American Scientists. [cited 2024 Mar 27]. Available from: https://fas.org/publication/bio-x-ai-policy- recommendations/

    work page 2024

  23. [23]

    arXiv.org

    AI4Science MR, Quantum MA. arXiv.org. 2023 [cited 2024 Mar 27]. The Impact of Large Language Models on Scientific Discovery: a Preliminary Study using GPT-4. Available from: https://arxiv.org/abs/2311.07361v2

    work page arXiv 2023

  24. [24]

    [cited 2024 Mar 27]

    Introducing Devin, the first AI software engineer [Internet]. [cited 2024 Mar 27]. Available from: https://www.cognition-labs.com/introducing-devin

    work page 2024

  25. [25]

    On the Opportunities and Risks of Foundation Models

    Bommasani R, Hudson DA, Adeli E, Altman R, Arora S, von Arx S, et al. On the Opportunities and Risks of Foundation Models [Internet]. arXiv; 2022 [cited 2024 Mar 31]. Available from: http://arxiv.org/abs/2108.07258

    work page internal anchor Pith review Pith/arXiv arXiv 2022

  26. [26]

    The White House

    House TW. The White House. 2023 [cited 2024 Mar 27]. Executive Order on the Safe, Secure, and Trustworthy Development and Use of Artificial Intelligence. Available from: https://www.whitehouse.gov/briefing-room/presidential-actions/2023/10/30/executive-order-on-the- safe-secure-and-trustworthy-development-and-use-of-artificial-intelligence/

    work page 2023

  27. [27]

    Sequence modeling and design from molecular to genome scale with Evo [Internet]

    Nguyen E, Poli M, Durrant MG, Thomas AW, Kang B, Sullivan J, et al. Sequence modeling and design from molecular to genome scale with Evo [Internet]. bioRxiv; 2024 [cited 2024 Mar 31]. p. 2024.02.27.582234. Available from: https://www.biorxiv.org/content/10.1101/2024.02.27.582234v2

    work page doi:10.1101/2024.02.27.582234v2 2024

  28. [28]

    Engineering AAVs with Evo and AlphaFold [Internet]

    Workman K. Engineering AAVs with Evo and AlphaFold [Internet]. 2024 [cited 2024 Mar 30]. Available from: https://blog.latch.bio/p/engineering-aavs-with-evo-and-alphafold

    work page 2024

  29. [29]

    Washington, D.C.: National Academies Press; 2006 [cited 2024 Mar 27]

    Globalization, Biosecurity, and the Future of the Life Sciences [Internet]. Washington, D.C.: National Academies Press; 2006 [cited 2024 Mar 27]. Available from: http://www.nap.edu/catalog/11567

    work page 2006

  30. [30]

    [cited 2024 Mar 27]

    Autonomous, multiproperty-driven molecular discovery: From predictions to measurements and back | Science [Internet]. [cited 2024 Mar 27]. Available from: https://www-science- org.laneproxy.stanford.edu/doi/10.1126/science.adi1407

    work page doi:10.1126/science.adi1407 2024

  31. [31]

    [cited 2024 May 6]

    Synthetic Nucleic Acids [Internet]. [cited 2024 May 6]. Available from: https://aspr.hhs.gov:443/legal/synna/Pages/default.aspx

    work page 2024

  32. [32]

    2024 [cited 2024 May 9]

    The White House [Internet]. 2024 [cited 2024 May 9]. Framework for Nucleic Acid Synthesis Screening | OSTP. Available from: https://www.whitehouse.gov/ostp/news- updates/2024/04/29/framework-for-nucleic-acid-synthesis-screening/

    work page 2024

  33. [33]

    Washington, D.C.: National Academies Press; 2018 [cited 2024 Mar 27]

    Biodefense in the Age of Synthetic Biology [Internet]. Washington, D.C.: National Academies Press; 2018 [cited 2024 Mar 27]. Available from: https://www.nap.edu/catalog/24890

    work page 2018

  34. [34]

    Systematic discovery of recombinases for efficient integration of large DNA sequences into the human genome

    Durrant MG, Fanton A, Tycko J, Hinks M, Chandrasekaran SS, Perry NT, et al. Systematic discovery of recombinases for efficient integration of large DNA sequences into the human genome. Nat Biotechnol. 2023 Apr;41(4):488–99

    work page 2023

  35. [35]

    [cited 2024 Mar 27]

    GOV.UK [Internet]. [cited 2024 Mar 27]. AI Safety Institute approach to evaluations. Available from: https://www.gov.uk/government/publications/ai-safety-institute-approach-to-evaluations/ai- safety-institute-approach-to-evaluations

    work page 2024

  36. [36]

    2024 [cited 2024 May 9]

    The White House [Internet]. 2024 [cited 2024 May 9]. United States Government Policy for Oversight of Dual Use Research of Concern and Pathogens with Enhanced Pandemic Potential | OSTP. Available from: https://www.whitehouse.gov/ostp/news-updates/2024/05/06/united-states- government-policy-for-oversight-of-dual-use-research-of-concern-and-pathogens-with-e...

    work page 2024

  37. [37]

    [cited 2024 Mar 27]

    GOV.UK [Internet]. [cited 2024 Mar 27]. Introducing the AI Safety Institute. Available from: https://www.gov.uk/government/publications/ai-safety-institute-overview/introducing-the-ai-safety- institute

    work page 2024

  38. [38]

    [cited 2024 Apr 2]

    GOV.UK [Internet]. [cited 2024 Apr 2]. The Bletchley Declaration by Countries Attending the AI Safety Summit, 1-2 November 2023. Available from: https://www.gov.uk/government/publications/ai- safety-summit-2023-the-bletchley-declaration/the-bletchley-declaration-by-countries-attending-the-ai- safety-summit-1-2-november-2023

    work page 2024

  39. [39]

    [cited 2024 Mar 27]

    Anthropic’s Responsible Scaling Policy [Internet]. [cited 2024 Mar 27]. Available from: https://www.anthropic.com/news/anthropics-responsible-scaling-policy

    work page 2024

  40. [40]

    [cited 2024 Mar 27]

    Preparedness [Internet]. [cited 2024 Mar 27]. Available from: https://openai.com/safety/preparedness

    work page 2024

  41. [41]

    [cited 2024 Mar 27]

    Introducing the next generation of Claude [Internet]. [cited 2024 Mar 27]. Available from: https://www.anthropic.com/news/claude-3-family

    work page 2024

  42. [42]

    Will releasing the weights of future large language models grant widespread access to pandemic agents? [Internet]

    Gopal A, Helm-Burger N, Justen L, Soice EH, Tzeng T, Jeyapragasan G, et al. Will releasing the weights of future large language models grant widespread access to pandemic agents? [Internet]. arXiv; 2023 [cited 2024 Mar 27]. Available from: http://arxiv.org/abs/2310.18233

    work page arXiv 2023

  43. [43]

    The WMDP Benchmark: Measuring and Reducing Malicious Use With Unlearning

    Li N, Pan A, Gopal A, Yue S, Berrios D, Gatti A, et al. The WMDP Benchmark: Measuring and Reducing Malicious Use With Unlearning [Internet]. arXiv; 2024 [cited 2024 Mar 27]. Available from: http://arxiv.org/abs/2403.03218

    work page internal anchor Pith review Pith/arXiv arXiv 2024

  44. [44]

    Measuring Massive Multitask Language Understanding

    Hendrycks D, Burns C, Basart S, Zou A, Mazeika M, Song D, et al. Measuring Massive Multitask Language Understanding [Internet]. arXiv; 2021 [cited 2024 Mar 27]. Available from: http://arxiv.org/abs/2009.03300

    work page internal anchor Pith review Pith/arXiv arXiv 2021

  45. [45]

    [Internet]

    Amazon Web Services, Inc. [Internet]. [cited 2024 Mar 27]. Responsible AI – AWS AI Service Cards: Amazon Titan Text – Amazon Web Services. Available from: https://aws.amazon.com/machine-learning/responsible-machine-learning/titan-text/

    work page 2024

  46. [46]

    Gemini: A Family of Highly Capable Multimodal Models

    Gemini Team, Anil R, Borgeaud S, Wu Y, Alayrac JB, Yu J, et al. Gemini: A Family of Highly Capable Multimodal Models [Internet]. arXiv; 2023 [cited 2024 Mar 27]. Available from: http://arxiv.org/abs/2312.11805

    work page internal anchor Pith review Pith/arXiv arXiv 2023

  47. [47]

    [cited 2024 Mar 27]

    Overview of Meta AI safety policies prepared for the UK AI Safety Summit | Transparency Center [Internet]. [cited 2024 Mar 27]. Available from: https://transparency.fb.com/en-gb/policies/ai- safety-policies-for-safety-summit/

    work page 2024

  48. [48]

    Microsoft On the Issues

    Blogs MC. Microsoft On the Issues. 2023 [cited 2024 Mar 27]. Microsoft’s AI Safety Policies. Available from: https://blogs.microsoft.com/on-the-issues/2023/10/26/microsofts-ai-safety-policies/

    work page 2023

  49. [49]

    [cited 2024 Jun 21]

    United States Government Policy for Oversight of Life Sciences DURC [Internet]. [cited 2024 Jun 21]. Available from: https://www.phe.gov/s3/dualuse/Pages/USGOversightPolicy.aspx

    work page 2024

  50. [50]

    [cited 2022 Nov 2]

    Department of Health and Human Services Framework for Guiding Funding Decisions about Proposed Research Involving Enhanced Potential Pandemic Pathogens (P3CO) [Internet]. [cited 2022 Nov 2]. Available from: https://www.phe.gov/s3/dualuse/Pages/ResearchReview-PPP.aspx

    work page 2022

  51. [51]

    Dual Use Research of Concern [Internet]

    admin. Dual Use Research of Concern [Internet]. Office of Science Policy. [cited 2021 Aug 16]. Available from: https://osp.od.nih.gov/biotechnology/dual-use-research-of-concern/

    work page 2021

  52. [52]

    Washington, D.C.: National Academies Press; 2004 [cited 2024 Mar 27]

    Biotechnology Research in an Age of Terrorism [Internet]. Washington, D.C.: National Academies Press; 2004 [cited 2024 Mar 27]. Available from: http://www.nap.edu/catalog/10827

    work page 2004

  53. [53]

    Governance of Dual Use Research in the Life Sciences: Advancing Global Consensus on Research Oversight: Proceedings of a Workshop [Internet]

    National Academies of Sciences E. Governance of Dual Use Research in the Life Sciences: Advancing Global Consensus on Research Oversight: Proceedings of a Workshop [Internet]. 2018 [cited 2021 Feb 18]. Available from: https://www.nap.edu/catalog/25154/governance-of-dual-use- research-in-the-life-sciences-advancing

    work page 2018

  54. [54]

    Public health and biosecurity

    Berns KI, Casadevall A, Cohen ML, Ehrlich SA, Enquist LW, Fitch JP, et al. Public health and biosecurity. Adaptations of avian flu virus are a cause for concern. Science. 2012 Feb 10;335(6069):660–1

    work page 2012

  55. [55]

    Protocols and risks: when less is more

    Pannu J, Sandbrink JB, Watson M, Palmer MJ, Relman DA. Protocols and risks: when less is more. Nat Protoc. 2022 Jan;17(1):1–2

    work page 2022

  56. [56]

    Generative Artificial Intelligence-Assisted Protein Design Must Consider Repurposing Potential

    Ekins S, Brackmann M, Invernizzi C, Lentzos F. Generative Artificial Intelligence-Assisted Protein Design Must Consider Repurposing Potential. GEN Biotechnol. 2023 Aug;2(4):296–300

    work page 2023

  57. [57]

    2023 [cited 2024 Mar 27]

    Federal Select Agent Program [Internet]. 2023 [cited 2024 Mar 27]. Available from: https://www.selectagents.gov/index.htm

    work page 2023

  58. [58]

    [cited 2024 Mar 27]

    Chemical and Biological Controls [Internet]. [cited 2024 Mar 27]. Available from: https://www.bis.doc.gov/index.php/policy-guidance/product-guidance/chemical-and-biological- controls

    work page 2024

  59. [59]

    [cited 2024 Mar 27]

    Australia Group Common Control Lists — The Australia Group [Internet]. [cited 2024 Mar 27]. Available from: https://www.dfat.gov.au/publications/minisite/theaustraliagroupnet/site/en/controllists.html

    work page 2024

  60. [60]

    Beyond Biosecurity by Taxonomic Lists: Lessons, Challenges, and Opportunities

    Millett P, Alexanian T, Brink KR, Carter SR, Diggans J, Palmer MJ, et al. Beyond Biosecurity by Taxonomic Lists: Lessons, Challenges, and Opportunities. Health Secur. 2023 Dec;21(6):521–9

    work page 2023

  61. [61]

    Proposed Biosecurity Oversight Framework for the Future of Science

    Letts K (NIH/OD) [C]. Proposed Biosecurity Oversight Framework for the Future of Science

    work page

  62. [62]

    Artificial Intelligence Safety Institute

    U.S. Artificial Intelligence Safety Institute. NIST [Internet]. 2023 Oct 26 [cited 2024 Mar 27]; Available from: https://www.nist.gov/artificial-intelligence/artificial-intelligence-safety-institute

    work page 2023